Run a review bundle on one note

Type: kb/types/instruction.md

Review a specific note against an explicit list of gates from inside the current agent harness.

Inputs:

  • first argument: {note-path} — repository-relative note path, for example kb/notes/backlinks.md
  • remaining arguments: {gate-or-bundle}... — one or more gate ids or bundle names, for example semantic/grounding-alignment, prose/source-residue, or prose (= all prose gates)

Do not run the selector to choose gates. Treat the provided list as the exact execution set.

Live agent path

1. Create the review run and canonical prompt

commonplace-create-review-run --runner {codex|claude-code} --model {model-id} --with-prompt {note-path} {gate-or-bundle}...

Capture the JSON output, especially:

  • review_run_id
  • prompt_path
  • bundle_output_path
  • gate_ids

Do not invent or reorder gate_ids; use exactly what the helper resolves.

2. Follow the canonical prompt

Read the file at prompt_path and treat it as the authoritative reviewer instruction. It contains the reading scope, gate definitions, link-resolution table, and sentinel output contract for this run.

Write the full sentinel-bracketed review bundle to bundle_output_path.

3. Ingest the bundle output

commonplace-ingest-bundle-output --review-run-id {review-run-id} --input-file {bundle-output-path}

This parses the bundle with the same parser used by commonplace-run-review-bundle, records the per-gate reviews, and finalizes the review run.

Do not

  • Do not run the selector to choose gates before reviewing. This instruction is for explicit execution.
  • Do not invoke commonplace-record-bundle-review; that command is no longer part of the shipped CLI.
  • Do not call commonplace-write-gate-review in the normal live-agent bundle flow; use commonplace-ingest-bundle-output instead.
  • Do not skip a requested gate block in the bundle output.